OpenAI's GPT-Live-1 Handles Simultaneous Speech, Ending Turn-Based Voice Delays

OpenAI released GPT-Live-1, a full-duplex voice model that processes incoming and outgoing audio concurrently, eliminating the turn-based wait that plagued earlier ChatGPT Voice. Rolling out now across iOS, Android, and web, the model handles mid-sentence interruptions, real-time translation, and verbal backchannels—"mhmm," "yeah," "got it"—that signal active listening. GPT-Live-1 routes complex queries to GPT-5.5 for reasoning. Paid subscribers access the full model; free users get GPT-Live-1 mini.